International Journal of Business and Finance Research

Vol. 9, No. 4, 2015, pp. 81-87

ISSN: 1931-0269 (print)
www.theIBFR.com
ISSN: 2157-0698 (online)

DETERMINANTS OF SILVER FUTURES PRICE

VOLATILITY: EVIDENCE FROM THE THAILAND
FUTURES EXCHANGE
Woradee Jongadsayakul, Kasetsart University

ABSTRACT

This research studies determinants of silver futures price volatility in Thailand Futures Exchange using
generalized autoregressive conditional heteroskedasticity model. The sample data consist of daily closing
price, volume, and open interest of silver futures from the period June 21, 2011 to December 26, 2012 for
the nearby month contract with 376 sample data points. I construct data sample by switching or rolling
over to the next maturing contract one day before the expiration date. The empirical results reveal there is
no significant relationship between volatility and time to expiration. There are a negative role for trading
volume and a positive role for open interest in determining silver futures price volatility. The analysis of
silver futures price volatility insists the Clearing House that margin requirements for silver futures should
not be affected as the time to maturity of the contract decreases. The findings are also helpful to risk
managers dealing with silver futures and predicting silver futures price volatility.

JEL: C32, G13, G32

KEYWORDS: Futures Price Volatility, Silver Futures, Samuelson Hypothesis

INTRODUCTION

D erivatives products such as forwards, futures, and options are great tools, which investors can use
to predict future spot prices and minimize their risk. Derivatives in Thailand have started long
before Thailand Futures Exchange (TFEX) came into existence. Usually they are in the form of an
over-the-counter between each counterparties. On May 17, 2004, TFEX, a subsidiary of the Stock Exchange
of Thailand (SET), was established as a derivatives exchange. TFEX has traded SET50 index futures,
SET50 index options, gold futures, silver futures, interest rate futures, single stock futures, crude oil futures,
USD futures, and sector futures since then. In 2013, TFEXs average daily volume was 68,017 contracts,
55.21% more than the 43,823 in the previous year. About 99.61% of the 2013 derivatives trading were
futures. Before trading futures in TFEX, investors are required to deposit initial margin with their respective
brokers to ensure that they fulfill their futures contract obligations. Initial margin requirements for futures
contracts are set by Thailand Clearing House Co., Ltd. (TCH). Their required rates are the same on the
same underlying across different maturities and typically only 10 to 15 percent of the full value of the
futures contracts. At the end of day, brokers will calculate the profit and loss and add or subtract funds via
a Mark-to-Market process. If the balance in the margin account falls below the maintenance margin level,
investor will receive a margin call to top up his or her margin account to meet the initial margin requirement.
One of the important factors affecting margin rate is futures price volatility. Therefore, understanding and
characterizing futures price volatility has been a key issue in futures market research. Previous research has
explained futures price volatility by variables such as time to maturity, volume, and open interest.

Samuelson (1965) states that volatility of futures prices should increase as the contract approaches
expiration. It is widely referred to as the Samuelson hypothesis. The logic behind this conclusion is that
the market is more sensitive to news regarding near-maturity contracts than more-distant contracts, which

81

Electronic copy available at: http://ssrn.com/abstract=2655734

W. Jongadsayakul | IJBFR Vol. 9 No. 4 2015

is indicated by greater volatility for the near-maturity contract (Ripple and Moosa, 2009). Numerous studies
have investigated the Samuelson hypothesis empirically, and the hypothesis has been supported in
commodity futures markets (Daal and Farhat, 2004; Duong and Kalev, 2008; Karali, Dorfman and
Thurman, 2009; Karali and Thurman, 2010) and currency futures markets (Madarassy Akin, 2003).
Samuelson hypothesis also holds in TFEX where SET50 futures price volatility (Dolsutham et al., 2011)
and gold futures price volatility (Jongadsayakul, 2014b) increase as expiry approaches. However, using
GARCH(1,1), Jongadsayakul (2014a) find the evidence of inverse maturity effect in case of crude oil
futures traded in TFEX. In addition, Chen, Duan and Hung (1999) find that the volatility of the Nikkei-225
index futures price decreases when the contract is closer to its maturity.

Futures trading activity proxied by volume and open interest is another important determinant of futures
price volatility. Numerous works have examined the relationship between trading volume and futures price
volatility. Considerable evidence exists a positive correlation between futures price volatility and trading
volume (Madarassy Akin, 2003; Xin, Chen and Firth, 2005; Kuo, Hsu and Chiang, 2005; Pati, 2006; Ripple
and Moosa, 2009; Jongadsayakul, 2014a, 2014b); however, Bessembinder and Seguin (1993) suggest that
the volatility-volume relationship might depend on the type of trader. Daigler and Wiley (1999) find that
the public drives the positive volatility-volume relationship whereas trades by clearing members and floor
traders often exhibit an inverse relationship between volatility and volume. Moreover, Avramov et al.
(2006) show that informed trades lead to a reduction in volatility while non-informational trades lead to an
increase in volatility. The introduction of open interest as an additional explanatory variable is motivated
by the fact that open interest and its change differ significantly from trading volume. The expectation is that
open interest is negatively related to volatility (Xin, Chen and Firth, 2005; Feng and Chuan-zhe, 2008;
Ripple and Moosa, 2009; Jongadsayakul, 2014b), as the availability of more contracts represents increased
market depth, implying greater liquidity. However, Madarassy Akin (2003) and Pati (2006) find the positive
relationship between open interest and financial futures price volatility. Higher open interest means that
there are more future trade expected and more opportunity for the prices to move into higher or lower levels.
This implies an increase in current futures price volatility.

This paper therefore examines the relationships between silver futures price volatility, time to expiration,
trading volume, and open interest in TFEX using Generalized Autoregressive Conditional
Heteroskedasticity (GARCH) model. The results from the MA(1,7,15)-GARCH(1,1) model reveal there is
no significant relationship between volatility and time to expiration. Silver futures price volatility is
negatively related to trading volume and positively related to open interest. The analysis of silver futures
price volatility insists TCH that margin requirements for silver futures should not be affected as the time to
maturity of the contract decreases. The findings are also helpful to risk managers dealing with silver futures
and predicting silver futures price volatility. Moreover, it adds to literature as another evidence of a negative
role for trading volume and a positive role for open interest in determining silver futures price volatility,
which are not as expected. The rest of the paper is organized as follows. In section 2, I discuss earlier
work related to determinants of futures price volatility. Then I describe the data and outline the methodology
in section 3. In section 4, I present empirical results from the MA(1,7,15)-GARCH(1,1) model. Finally, I
conclude the paper and summarize the findings in section 5.

LITERATURE REVIEW

The literature contains numerous studies attempting to identify the important variables that influence futures
price volatility. Samuelson (1965) states that volatility of futures prices should increase as the contract
approaches expiration. It is widely known as the Samuelson hypothesis. His study is based on two
assumptions, namely that: (1) futures price equals expected future spot price, and (2) the spot price is
stationary following an AR(1): = 1 + , where ( ) = 0 and (2 ) = 2 . Bessembinder et al.
(1996) presents a new analysis of the economic issues underlying the prediction of greater volatility for the
near-maturity futures contract. They show the Samuelson hypothesis is successful without the two

82

Electronic copy available at: http://ssrn.com/abstract=2655734

The International Journal of Business and Finance Research VOLUME 9 NUMBER 4 2015

assumptions. In their study, the change in the log futures price ( ) is the summation of the ex-ante spot
market risk premium (), the unexpected rate of spot market price appreciation ( ), and the change in the
futures term slope weighted by the remaining time to expiration ( ). Therefore, the variance of futures
price changes is

VAR( ) = VAR( ) + 2 VAR( ) + 2COV( , ). (1)

where VAR and COV denote variance and covariance, respectively. If VAR( ) is a positive constant,
then the variance of futures price increases with the square of the remaining time to contract expiration.
Higher variance for more distant delivery date is inconsistent with the Samuelson hypothesis. Unlike
VAR( ), the sign of COV( , ) can be negative or positive. If COV( , ) < 0, then the third term
on the right side of (1) increases (toward zero) as the trading date approaches the contract expiration date.
Therefore, their analysis implies that the Samuelson hypothesis requires negative covariance between spot
returns and the slope of the futures term structure, which occurs in markets where prices are mean reverting.

The Samuelson hypothesis has been empirically supported in commodity futures markets (Daal and Farhat,
2004; Duong and Kalev, 2008; Karali, Dorfman and Thurman, 2009; Karali and Thurman, 2010) and
currency futures markets (Madarassy Akin, 2003). Samuelson hypothesis also holds in cases of SET50
futures price volatility (Dolsutham et al., 2011) and gold futures price volatility (Jongadsayakul, 2014b).
However, Chen, Duan and Hung (1999) find the inverse maturity effect in the Nikkei-225 index futures
price volatility. Other studies also show that their results depend on the data and the methodology. Floros
and Vougas (2006) find that the Samuelson hypothesis seems to be correct in simple linear regressions for
both FTSE/ASE-20 and FTSE/ASE Mid 40 index. On the other hand, using GARCH models, there is no
evidence to support Samuelsons hypothesis in FTSE/ASE-20 index over the whole period. However, using
monthly series, GARCH models show a stronger support to the Samuelson hypothesis rather than linear
regressions. Moreover, Jongadsayakul (2014a) shows that there is no significant relationship between crude
oil futures price volatility and time to expiration in linear regression model while there is an evidence of
inverse maturity effect in GARCH(1,1) model.

Futures trading activity proxied by volume and open interest is another important determinant of futures
price volatility. When trading volume increases, it increases the probability that prices will move into higher
or lower regions, and that their volatility will be greater than before (Xin, Chen and Firth, 2005). On the
other hand, trading volume can be used as a measure of the information flow. Therefore, when new and
unexpected information arrives, both volatility and volume change contemporaneously and positively to
new information. Numerous works identify a strong positive relation between price volatility and trading
volumes (Madarassy Akin, 2003; Xin, Chen and Firth, 2005; Kuo, Hsu and Chiang, 2005; Pati, 2006;
Ripple and Moosa, 2009; Jongadsayakul, 2014a, 2014b). However, Bessembinder and Seguin (1993)
suggest that the volatility-volume relationship might depend on the type of trader.

Daigler and Wiley (1999) find that the public drives the positive volatility-volume relationship whereas
trades by clearing members and floor traders often exhibit an inverse relationship between volatility and
volume. Moreover, Avramov et al. (2006) show that informed trades lead to a reduction in volatility while
non-informational trades lead to an increase in volatility. Besides trading volume, open interest is important
indicator of trading activity. It reflects the current willingness of futures traders to risk their capital in the
futures position, which indicates the level of market depth (Bessembinder and Seguin, 1993). A high level
of open interest could help to create market conditions that would reduce pressure from prices when trading
provides new information (Xin, Chen and Firth, 2005). Several studies find that open interest is negatively
related to volatility (Xin, Chen and Firth, 2005; Feng and Chuan-zhe, 2008; Ripple and Moosa, 2009;
Jongadsayakul, 2014b). However, Madarassy Akin (2003) and Pati (2006) find the positive relationship
between open interest and financial futures price volatility. Higher open interest means that there are more

83
W. Jongadsayakul | IJBFR Vol. 9 No. 4 2015

future trade expected and more opportunity for the prices to move into higher or lower levels. This implies
an increase in current futures price volatility.

DATA AND METHODOLOGY

This study examines determinants of silver futures price volatility by using daily data downloaded from the
websites of SETSMART and TFEX. The sample data consist of daily closing price, time to maturity, trading
volume, and open interest of silver futures from the period June 21, 2011 to December 26, 2012 for the
nearby month contract with 376 sample data points. I construct data sample by switching or rolling over to
the next maturing contract one day before the expiration date. To analyze the volatility of silver futures
price, I employ Generalized Autoregressive Conditional Heteroskedasticity (GARCH) methodology. Based
on Akaike Information Criterion (AIC) and Schwarz-Bayes Criterion (SBC), a MA(1,7,15)-GARCH(1,1)
model is chosen. Moreover, the basic GARCH specification is augmented by time to maturity variable and
two trading activity variables, open interest and trading volume, in order to determine their relative
contribution to the conditional variance. The following model is then estimated to investigate silver futures
price volatility in TFEX.

= 0 + + 1 1 + 7 7 + 15 15 ; |1 ~(0, ) (2)
2
= 0 + 1 1 + 1 1 + 1 + 2 + 3 (3)

where Rt is the futures return, which are obtained by taking the difference of log of futures prices, ht is the
conditional variance, mt is the time to maturity, qt is the trading volume, ot is the open interest, 0, 0, i,
and bi are parameters, 1 is the ARCH coefficient, 1 is the GARCH coefficient. The sum of ARCH and
GARCH coefficients indicates the degree of persistent in volatility.

RESULTS

To study determinants of silver futures price volatility, the MA(1,7,15)-GARCH(1,1) model is augmented
by including time to maturity (m), trading volume (q), and open interest (o) as explanatory variables in the
conditional variance equation. Then Bellerslev-Woodbridges Quasi-Maximum Likelihood (QML) is used
to estimate the MA(1,7,15)-GARCH(1,1) model. Table 1 presents the estimated results of MA(1,7,15)-
GARCH(1,1) model, displaying the estimated coefficients and their P-values, as well as diagnostics tests
and Wald test. The validity of the estimated model is assessed first by testing the standardized residuals for
the presence of serial correlation and heteroskedasticity. Ljung-Box Q-test statistics up to lags 36 for serial
correlation in the standardized residuals and standardized squared residuals are 18.224 (Prob. = 0.9824) and
38.456 (Prob. = 0.2361), respectively. Then a Lagrange Multiplier test is employed to examine whether the
standardized squared residuals exhibit additional ARCH. LM ARCH (5) test is 8.272 (Prob. = 0.1419),
which implies that the null hypothesis of no ARCH effects cannot be rejected and reveals that no further
ARCH effects left in the standardized residuals.

The insignificant Ljung-Box Q statistics and LM ARCH statistics implies that the residuals of the estimated
MA(1,7,15)-GARCH(1,1) model are reasonably well behaved and adequately capture the persistence in the
variance of returns. Moreover, the estimation results of MA(1,7,15)-GARCH(1,1) model show that the
coefficient for the previous shock (the ARCH coefficient: 1 ) is 0.1787 and that for its lagged conditional
variance (the GARCH coefficient: 1 ) is 0.5080. The GARCH coefficient (1 ) is highly statistically
significant at the level of 1%. The sum of the two coefficients is 0.6867, which implies that the persistent
in volatility is not so high. Since the Wald test shows the sum of the two coefficients statistically less than
1 at the 1% significance level, the volatility process is co-variance stationary, stable and evidences of mean
reverting. Maturity is shown to be insignificant. The estimated coefficients of volume and open interest are
-0.0003 and 0.0003, respectively, which are statistically significant at the level of 10% and 5%, respectively.

84
The International Journal of Business and Finance Research VOLUME 9 NUMBER 4 2015

However, the signs of trading volume and open interest coefficients are not as expected. The results indicate
a negative relationship between trading volume and volatility due to higher informed trades in TFEX. An
increase in open interest means that there are more future trade expected and more opportunity for the prices
to move into higher or lower levels. This leads to an increase in current futures price volatility.

= 0 + + 1 1 + 7 7 + 15 15 ; |1 ~(0, )
2
= 0 + 1 1 + 1 1 + 1 + 2 + 3

Table 1: Estimation Results of MA(1,7,15)-GARCH(1,1) Model

Conditional Mean Estimated Value P-Value

Equation
Coefficient/Statistic

0 -0.0005 0.4871

1 -0.0759 0.1998

7 0.0592 0.2481

15 -0.1051** 0.0459

Conditional Variance Equation

0 0.0001 0.1630

1 0.1787 0.2225

1 0.5080*** 0.0092

1 0.0000 0.7573

2 -0.0003* 0.0651

3 0.0003** 0.0378

Standardized Residual Diagnostics

Ljung-Box Q (36) 18.224 0.9824

Ljung-Box Q2 (36) 38.456 0.2361

LM ARCH (5) 8.272 0.1419

Wald Test

H0: 1 + 1 1 -2.490*** 0.0064

This table shows the estimation results of MA(1,7,15)-GARCH(1,1) Model. Diagnostics tests based on LM ARCH (5), Ljung-Box Q (36), and Ljung-
Box Q2 (36) of residual and squared residual indicate that the estimated model is well fit. The sum of ARCH and GARCH coefficients is 0.6867,
indicating that the persistent in volatility is not so high. Wald test shows its value statistically less than one, indicating that volatility process is co-
variance stationary, stable and evidences of mean reverting. Maturity is insignificant, whereas trading volume and open interest tend to be
significant. ***, **, and * indicate statistically significant at the 1%, 5%, and 10% level, respectively.

CONCLUDING COMMENTS

This research studies determinants of silver futures price volatility in Thailand Futures Exchange using the
MA(1,7,15)-GARCH(1,1) model. The sample data consist of daily closing price, volume, and open interest
of silver futures from the period June 21, 2011 to December 26, 2012 for the nearby month contract with
376 sample data points. I construct data sample by switching or rolling over to the next maturing contract
one day before the expiration date. The results from the MA(1,7,15)-GARCH(1,1) model reveal there is no
significant relationship between silver futures price volatility and time to expiration, whereas trading
volume and open interest tend to be significant. However, the signs of trading volume and open interest

85
W. Jongadsayakul | IJBFR Vol. 9 No. 4 2015

coefficients are not as expected. The results indicate a negative relationship between silver futures price
volatility and trading volume, and a positive relationship between silver futures price volatility and open
interest. Therefore, trading volume and open interest are the two important variables, explaining the price
volatility of silver futures traded in TFEX. Since margin requirement is affected by futures price volatility,
the results of this study will assist TCH in setting margin requirements. No consistent relationship between
silver futures price volatility and time-to-maturity indicates that the margin requirement should not be
affected as the time-to-maturity of the silver futures contract decreases. The findings are also helpful to risk
managers dealing with silver futures and predicting silver futures price volatility. It is noted that the role of
trading volume on silver futures price volatility is negative. The inverse relationship between volume and
volatility is possible due to trading by clearing members and floor traders. However, this study uses total
volume. Therefore, further analysis of this issue should include both informed trades and non-informational
trades to confirm the results.

REFERENCES

Avramov, D., T. Chordia and A. Goyal (2006) The Impact of Trades on Daily Volatility, Review of
Financial Studies, vol. 19(4), p. 1241-1277.

Bessembinder, H., J.F. Coughenour, P.J. Seguin and M.M. Smeller (1996) Is There a Term Structure of
Futures Volatilities? Reevaluating the Samuelson Hypothesis, The Journal of Derivatives, vol. 4(2), p.
45-58.

Bessembinder, H. and P.J. Seguin (1993) Price Volatility, Trading Volume, and Market Depth: Evidence
from Futures Markets, Journal of Financial and Quantitative Analysis, vol. 28(1), p. 21-39.

Chen, Y.-J., J.-C. Duan and M.-W. Hung (1999) Volatility and Maturity Effects in the Nikkei Index
Futures, The Journal of Futures Markets, vol. 19(8), p. 895-909.

Daal, E. and J. Farhat (2004) Does Futures Exhibit Maturity Effect? New Evidence from an Extensive
Set of US and Foreign Futures Contracts, Retrieved from http://scholarworks.uno.edu/econ_wp/22.

Daigler, R.T. and M.K. Wiley (1999) The Impact of Trader Type on the Futures Volatility-Volume
Relation, Journal of Finance, vol. 54(6), p. 2297-2316.

Dolsutham, O., P. Payattikul, S. Yanwaree and P. Tharavanij (2011) Volatility of SET50 Index Futures
and Samuelson Hypothesis, Nida Business Journal, vol. 9, p. 71-103. (In Thai)

Duong, H.N. and P.S. Kalev (2008) The Samuelson Hypothesis in Futures Markets: An Analysis Using
Intraday Data, Journal of Banking & Finance, vol. 32(4), p. 489-500.

Feng, W. and L. Chuan-zhe (2008) Determinants of the Volatility of Futures Markets Price Returns: The
Case of Chinese Wheat Futures, International Conference on Management Science & Engineering
(15th). Long Beach, CA.

Floros, C. and D.V. Vougas (2006) Samuelsons Hypothesis in Greek Stock Index Futures Market,
Investment Management and Financial Innovations, vol. 3(2), p. 154-170.

Jongadsayakul, W. (2014a) Determinants of the Crude Oil Futures Price Volatility: The Case of
Thailand Futures Exchange. Proceeding of 52nd Kasetsart University Annual Conference, Bangkok
Thailand. (In Thai)

86
The International Journal of Business and Finance Research VOLUME 9 NUMBER 4 2015

Jongadsayakul, W. (2014b) "Determinants of the Gold Futures Price Volatility: The Case of Thailand
Futures Exchange," Applied Economics Journal, vol. 21(1), p. 59-78. (In Thai)

Karali, B., J.H. Dorfman and W.N. Thurman (2009) Does Futures Price Volatility Differ Across
Delivery Horizon, Proceedings of the NCCC-134 Conference on Applied Commodity Price Analysis,
Forecasting, and Market Risk Management. St. Louis, MO.

Karali, B. and W.N. Thurman (2010) Components of Grain Futures Price Volatility, Journal of
Agricultural and Resource Economics, vol. 35(2), p. 167-182.

Kuo, W.-H., H. Hsu and C.-Y. Chiang (2005) Price Volatility, Trading Activity and Market Depth:
Evidence from Taiwan and Singapore Taiwan Stock Index Futures Markets, Asia Pacific Management
Review, vol. 10(1), p. 131-143.

Madarassy Akin, R. (2003) Maturity Effects in Futures Markets: Evidence from Eleven Financial
Futures Markets, UC Santa Cruz Economics Working Paper No. 03-6.

Pati, P.C. (2006) Maturity and Volume Effects on the Volatility: Evidences from NSE Fifty Futures,
10th Capital Markets Conference, Indian Institute of Capital Markets Paper.

Ripple, R.D. and I.A. Moosa (2009) The Effects of Maturity, Trading Volume, and Open Interest on
Crude Oil Futures Price Range-Based Volatility, Global Finance Journal, vol. 20(3), p. 209-219.

Samuelson, P.A. (1965) Proof That Properly Anticipated Prices Fluctuate Randomly, Industrial
Management Review, vol. 6(2), p. 41-49.

Xin, Y., G. Chen and M. Firth (2005) The Determinants of Price Volatility in China's Commodity
Futures Markets, China Accounting and Finance Review, vol. 7(1), p. 124-145.

ACKNOWLEDGMENTS

The author acknowledges financial support from Department of Economics, Kasetsart University, and the
helpful comments of the journal editors, Terrance Jalbert and Mercedes Jalbert, and two anonymous
reviewers.

BIOGRAPHY

Woradee Jongadsayakul is Assistant Professor of Economics at Kasetsart University. Her research focuses
on Financial Economics and Project Evaluation. She can be reached at Faculty of Economics, Kasetsart
University, 50 Ngam Wong Wan Road, Chatuchak, Bangkok 10900 Thailand, woradee.j@ku.ac.th.

87